The Direction Is Clear

The Next Stage of WINHOTEL

The direction is clear.

The tools have been chosen.

The architecture has been defined.

There is a plan again.

The past few years have been marked by major technological changes and new possibilities. Many approaches were explored, evaluated and, in some cases, set aside. With the decision to build an AI-First development environment, our future direction is now defined.

We know which tools we want to use, which systems we want to keep, and how we intend to develop WINHOTEL in the years ahead.

This document describes the foundation of that new path.

Preface

Software development is currently undergoing a period of significant change. This affects desktop applications, web applications, traditional manual programming and, above all, the growing collaboration between developers and artificial intelligence.

From the perspective of a developer who experienced the transition from DOS to Windows, some familiar patterns can be recognized. New tools emerge while existing systems continue to be used. For a time, both worlds coexist. Eventually, however, there comes a point when the new tools become a natural and permanent part of everyday development work.

Within our company, our existing development and production systems will remain in place. They have proven themselves over many years and continue to provide the foundation for our customer installations.

At the same time, we are building a new development environment designed from the ground up for collaboration with artificial intelligence. Our goal is not to replace what already works, but to extend and improve it where it makes sense.

We follow a number of guiding principles:

• We rely on standardized server installations.
• We use self-hosting for our web applications.
• We protect our systems using Windows Server Backup combined with Synology backup solutions.
• We keep only the components required for the current WINHOTEL release.
• We treat artificial intelligence as a core element of future software development.

This document describes the setup of this new AI-First Development Environment and serves as the foundation for the next stage of WINHOTEL.

As the project evolves, we will document the development workstation, key decisions, tools, processes and lessons learned in an ongoing and informal manner. This document is intended to grow over time and serve both as a technical handbook and as a record of the project's evolution.

2026-06-12

Why This Document Exists

The first part of this document, up to and including Section 5, describes the standard WINHOTEL server platform.

This section is intentionally included even though it is largely unrelated to software development. The reason is simple: every development server is built on exactly the same foundation as every production server.

Only when all systems share the same structure, security configuration, network setup, backup strategy, and maintenance procedures can software be developed, tested, and deployed reliably.

The server foundation is therefore not the main subject of this document. It merely defines the technical baseline on which all further work is performed.

The actual focus of this ServerBook begins with Section 6.

From this point onward, the document describes how a standard Windows server is transformed into a modern AI-First development environment for WINHOTEL.

The goal is no longer simply to operate a server, but to create a development platform on which human developers and AI agents can work together efficiently.

Topics covered from Section 6 onward include:

• Development infrastructure
• Web development environment
• Harbour and FIVEWIN toolchains
• Agent-based software development
• Prompt-driven engineering
• Knowledge management
• AI skills and coding standards
• Integration of ChatGPT, Claude Code, Codex, and future AI systems

The result is a standardized WINHOTEL Development Server that is technically compatible with production systems while simultaneously serving as a platform for AI-assisted software engineering.

In short:

Sections 1–5 describe the server.

Sections 6 onward describe how that server becomes an AI-First WINHOTEL development workstation.

WINHOTEL SERVERBOOK

Standardized Installation of a WINHOTEL Server and Development Server

Objective

This document describes the standardized setup of a WINHOTEL server.

Every server is built according to a strictly defined procedure. Individual configurations are avoided. All systems should have the same structure, the same security settings, and the same maintenance mechanisms.

This creates a reproducible production system rather than a custom one-off installation.

Phase 1 – Basic Server Installation

1. Prepare Hardware

The basis is a standard notebook or PC intended to be used as a WINHOTEL server.

Example

• HP ProBook
• Windows 11 preinstalled

Before reinstallation, the existing device drivers are exported.

Back Up Drivers

All Windows 11 drivers are exported and archived.

Purpose:

• Protection against missing manufacturer drivers
• Reinstallation of unrecognized devices under Windows Server

2. Install Windows Server 2025

Install

• Windows Server 2025 Standard

Basic Configuration

• Assign computer name
• Set administrator password
• Configure network settings

After Completion

• Run Windows Update
• Restart the server

3. Install Missing Drivers

Check Device Manager.

If devices are missing:

• Import drivers from the Windows 11 driver export
• Recheck the device

Goal:

• No unknown devices
• Fully functional hardware

Phase 2 – WINHOTEL Server Foundation

4. Run the Installation USB Stick

The WINHOTEL installation USB stick contains all standard scripts.

The installation is menu-driven.

The script automatically configures:

System Configuration

• Power options
• Disable hibernation
• Configure automatic restarts
• Time synchronization

Users and Security

• Enable RDP
• Create RDP user groups
• Create standard folders
• Create shared folders

Firewall

• Standard rules
• WINHOTEL rules
• RDP rules
• Backup rules

Backup

• Windows Server Backup
• Synology Active Backup
• Backup schedules

Maintenance

• Event logging
• Monitoring foundation
• Standard services

After completion, the server conforms to the WINHOTEL standard.

Phase 3 – WINHOTEL Installation

5. Install WINHOTEL Desktop Applications

Installation is performed exclusively through the existing installation scripts.

The following are installed:

• WINHOTEL core system
• Required modules
• Printing components
• Additional utilities

Manual installations should be avoided.

6. Network and Internet Services

Install and configure:

• Internet connectivity
• DNS
• Domain services
• SSL certificates
• Apache web server
• PHP
• Required runtime environments

Setup is performed through the available script menus.

Phase 4 – System Verification

7. Server Testing

The following items must be successfully verified:

Network

• Server is reachable
• Internet access is available
• DNS is functioning

Remote Desktop

• RDP login is possible
• User permissions are correct

WINHOTEL

• Application starts successfully
• Data access works
• User login functions correctly

Web Services

• Apache is reachable
• HTTPS is reachable
• Domain is reachable

Backup

• Synology connection established
• Backup job configured
• Test run completed successfully

Phase 5 – WINHOTEL Online Modules

8. Verify Online Modules

Check all production modules in use:

• Online booking
• Channel manager
• Interfaces
• APIs
• Web modules

The installation is considered complete only after successful verification.

Phase 6 – Development Server

9. Extend to a Development Server

In addition to the production environment, a development environment is installed.

The development server is used for developing future WINHOTEL versions.

10. FIVEWIN / Harbour Development Environment

Install

• Harbour
• FIVEWIN
• Compiler
• Development tools
• Build scripts

Only the components required for the current WINHOTEL release are installed.

Unused legacy components are intentionally removed.

Goal:

• Clear development environment
• Reduced complexity
• Easier maintenance

11. Web Development Environment

Install

• Visual Studio Code
• Node.js LTS
• Apache
• PHP
• PE_EDITOR
• WebsiteBuilder

Project Structure

C:\www\htdocs
├─ PE_EDITOR
├─ websitebuilder
├─ winhotel\

The development server remains compatible with the production environment.

Phase 7 – AI Development First

12. AI-Assisted Development

The development server is prepared for agentic programming.

Systems Used

• ChatGPT (general-purpose AI)
• Claude Code
• OpenAI Codex CLI
• Additional models as required

AI is not used directly on production systems.

Development takes place exclusively on the development server.

13. Prompt Infrastructure

The foundation of development:

PromptBuilder

Defines:

• What should be built

PE_EDITOR

Manages:

• Skills
• Knowledge base
• Project profiles
• Agent exports

Knowledge Base

Contains:

• WINHOTEL rules
• Architectural knowledge
• Project experience

Skills

Define:

• AI working methods
• Programming rules
• Quality requirements

Target State

The final result is a standardized WINHOTEL server with the same configuration as all production systems, including an integrated AI development environment for the continued development of future WINHOTEL versions.

Migration of the FIVEWIN/Harbour environment – I do not want to copy over things that are no longer relevant.

I don’t know how you manage your projects, but I have various folders and have now searched through all of them, discovering that more than 300 projects have accumulated over the past 30+ years. Here is an initial overview.

I now need to get an overview of what is no longer current and what were merely test projects.

I am not deleting anything, but I will only migrate what is actually required for the releases.

Additionally, I have now created an inventory program. As a side benefit, the inventory program generates not only the *.prj files but also batch files for compiling and linking. This means I will no longer need my old editor and will only take VS Code with me.

Excuse me Mr. Otto, but what does all this have to do with FiveWin and Harbour?

Hello Silvio,

the connection to FiveWin and Harbour is actually quite direct.

This is not primarily a document about server administration. It is about how to move an existing FiveWin/Harbour development environment to a new development server in a structured and reproducible way.

Specifically, it covers:

Installation and standardization of the FiveWin/Harbour toolchain Migration of existing projects Reviewing and organizing more than 300 accumulated projects Keeping only the components required for current releases Generating .prj files and build scripts automatically Moving development to VS Code Preparing the environment for AI-assisted and agentic software development

The first part intentionally describes the server foundation because our development servers and production servers are built on the same standardized platform. Only with an identical infrastructure can development, testing, and deployment remain reliable.

The actual FiveWin/Harbour development topics begin in Section 6, where the development environment, project migration, build process, and future development workflow are described.

In short, this document is about the next-generation development environment for a large existing FiveWin/Harbour application, not simply about server installation.

Best regards, Otto

Current Status

The development environment has now been fully prepared.

The hardware has been set up, and the existing project files have been copied to the server. A particularly positive aspect is that both SiteWheel and the hardware required virtually no additional configuration. As a result, the environment became operational very quickly.

In addition, the entire AI toolchain was installed automatically through scripts.

The following components have been successfully installed and verified:

Current Status

The WINHOTEL AI-First development server has been successfully deployed and verified.

The hardware was prepared, project files were transferred, and the development environment became operational with almost no manual configuration. SiteWheel and the underlying hardware required virtually no additional setup.

The complete AI toolchain was installed and tested successfully.

Installed Components

• Visual Studio Code 1.124.2
• Node.js 24.16.0
• npm 11.13.0
• Git 2.54.0
• Google Chrome
• OpenAI Codex CLI 0.139.0
• Claude Code 2.1.177
• Gemini CLI 0.46.0
• Google Antigravity Desktop

Development Structure

C:\Projects
├─ WINHOTEL
├─ WebsiteBuilder
├─ PE_EDITOR
├─ Skills
└─ KnowledgeBase

Verification Results

All major development tools were successfully installed and verified. The project structure was created automatically, and the required PATH configuration was completed successfully.

Result

The server is now fully operational and ready for AI-First software development using Codex, Claude Code, Gemini CLI, and Google Antigravity within a standardized WINHOTEL development environment.

Is This Already an Agentic Development System?

The short answer is: yes, partially.

What we have built so far is already much more than a collection of prompts. It contains several of the core building blocks of an agentic development system.

Traditional Prompt-Based Workflow

A traditional AI workflow looks like this:

User Request
→ AI Model
→ Response

Each conversation starts with little or no persistent context.

Current WINHOTEL AI-First Architecture

Our current structure already introduces persistent context and reusable knowledge:

Knowledge Base
→ Skills
→ Project Profiles
→ PromptBuilder
→ PE_EDITOR
→ AI Provider (Claude, Codex, Gemini, ChatGPT)

Instead of interacting directly with the AI model, requests are enriched with project knowledge, development rules and contextual information before being sent.

This is already a simple form of an agentic architecture.

What Is Still Missing?

1. Persistent Memory

The system should remember decisions and project status across sessions.

Examples:

• What was the last architectural decision?
• Which tasks are still open?
• What is the next planned step?

A possible structure could be:

Projects
└─ PE_EDITOR
├─ Memory
│ ├─ current_state.md
│ ├─ open_tasks.md
│ └─ decisions.md

This would allow future AI sessions to continue from the current project state.

2. Task Management

Today:

Request
→ AI Response

Agentic approach:

Request
→ Analysis
→ Task Breakdown
→ Execution
→ Status Update

The system should track work in progress and planned next steps.

3. Automatic Project Context

The development environment should automatically recognize the current project and load:

• Knowledge Base
• Skills
• Architecture Notes
• Memory
• Open Tasks

without requiring manual preparation every time.

4. Tool Usage

Modern AI coding tools already provide access to:

• File reading
• File editing
• Code analysis
• Project search

This gives them agent-like capabilities when working inside a project.

An Important Observation

The actual agent is not Claude, Codex, Gemini or ChatGPT.

The real agent is:

PE_EDITOR

The AI models are interchangeable execution engines.

The architecture can be viewed as:

WINHOTEL Knowledge
↓
WINHOTEL Skills
↓
Project Profiles
↓
PE_EDITOR
↓
AI Provider
(Claude / Codex / Gemini / ChatGPT)

This means the knowledge belongs to WINHOTEL, not to any specific AI vendor.

Models can change over time.

The knowledge, rules and development process remain under our control.

Next Logical Step

The next evolution of the system would be to introduce four core areas:

• Knowledge Base
• Skills
• Projects
• Memory

The final prompt would then be assembled from:

Knowledge

• Skill
• Project Profile
• Memory
• Current Task

At that point, the system would move beyond prompt management and become a true agentic development platform for WINHOTEL.

In this model, PE_EDITOR acts as the central orchestration layer, while different AI models remain replaceable execution engines behind it.

Dear Otto,

What is PE_EDITOR ?

Dear Antonio,
please see here: https://forums.fivetechsupport.com/viewtopic.php?t=46576

Best regards,
Otto

Dear Otto,

In a true Agentic system we have tools and skills.

The Agentic system may use any tool and/or skill, and even more, may be able to build new skills and tools upon demand at runtime.

We will review all these concepts next friday on the Agentic AI webinar :idea:
Friday 19 at 19 pm (spanish time)

Class Agent

https://fivetechsoft.github.io/FWH_docs/en/ai/agent.html

Dear Antonio,

thank you for your comments. I think we may actually be looking at two different layers of the same problem.

I fully agree that a true agentic system consists of skills and tools, and your Agent class is a very interesting implementation of that idea.

My current thinking comes from a slightly different direction.

Over the last months I realized that many of the technical problems I face every day are no longer the difficult part. Modern systems such as Codex, Claude Code and Gemini CLI already provide planning, repository analysis, file operations, shell execution and many other agent-like capabilities out of the box.

For FIVEWIN/Harbour application development, PHP, JavaScript, HTML, web applications and many everyday programming tasks, the models already provide a surprisingly strong knowledge base through their training data.

Because of that, I am currently asking myself a different question:

Where should I invest my own development effort?

At the moment I am less interested in building yet another execution engine and more interested in preserving and organizing project knowledge, workflows, company know-how, decisions, skills and working methods.

The PE_EDITOR is therefore evolving in a direction that is perhaps broader than software development alone.

The goal is to support everyday work across an entire small company:

• software development
• web development
• documentation
• support
• project management
• business processes
• knowledge management

My current architecture is based on:

Knowledge Base
→ Skills
→ Project Profiles
→ Memory
→ Agent Adapters
→ AI Provider

The AI provider itself is intentionally replaceable.

Another practical consideration is cost.

Whenever possible, I prefer using systems that work within existing subscriptions instead of requiring additional API usage. Not because APIs are bad, but because a fixed monthly cost lowers the barrier to everyday use. If every task requires asking, "Is this worth spending API credits on?", people tend to use the tools less often.

So my current assumption is:

Let Codex, Claude Code, Gemini and similar systems provide the execution layer.

Let the PE_EDITOR focus on the knowledge layer.

In my view, the long-term value lies less in the execution engine itself and more in preserving project knowledge, reusable skills, workflows, decisions and company-specific know-how independently from any particular AI provider.

Of course, I may be completely wrong.

One possible criticism of my approach is that I am delegating too much of the actual agentic behavior to external systems and keeping too little inside my own platform.

I would therefore be very interested in hearing where you think this approach is weak, where a dedicated agent framework provides significant advantages, and which parts you would still keep under your own control even in a world where Codex, Claude Code and similar tools continue to become more capable every month.

Best regards,

Otto

Dear Otto,

I needed weeks to fully understand the Agentic model.

With such understading I have been able to implement new development tools.

With those tools I plan to develop apps in a different way.

How a truly Agent AI system may look and behave like ?

Still I have more questions than answers :)

FiveTech Agents:

https://github.com/FiveTechSoft/Agents